Detailed methodological information

Here you will find specific information on the use of data for education monitoring studies.

Find studies

back to the study search

How many classes per school are included in the sample in the IQB studies?

In the IQB studies, one class per school is usually included in the sample. Exceptions are made for some federal states and for some types of schools (e.g. special education schools). Information on sampling in the studies can be found in the results reports or scale manuals.

Here is a brief summary of the sampling procedure:

  • National Assessment Study 2008/2009: One 9th grade class per school; the entire class took part in the test; special education schools were not part of the sample.
  • National Assessment Study 2011: in regular schools: One 4th grade class per school; the entire class took part in the test; at special schools, all students in 4th grade with a special need in the area of learning, language, or emotional and social development participated across all classes.
  • National Assessment Study 2012: In grammar schools ("Gymnasium"), one 9th grade class was included in the study, in other school types (with the exception of special education schools), two classes per school (if available) were included. The entire classes took part in the test. At special schools, all students in 4th grade with a special need in the area of learning, language, or emotional and social development participated across all classes.
  • IQB Trends in Student Achievement 2015: In regular schools, one ninth grade class per school was included in the sample; the entire class took part in the test. In special education schools, all ninth grade adolescents with special needs in the area of learning, language, or emotional and social development participated in the study.
  • IQB Trends in Student Achievement 2016: in regular schools: one 4th grade class per school; the entire class took part in the test; at special schools, all students in 4th grade with a special need in the area of learning, language, or emotional and social development participated across all classes.
  • IQB Trends in Student Achievement 2018: In grammar schools ("Gymnasium"), one 9th grade class was included in the study, in other school types (with the exception of special education schools), two classes per school (if available) were included. The entire classes took part in the test. At special schools, all students in 4th grade with a special need in the area of learning, language, or emotional and social development participated across all classes.

Are the competence estimators of the PISA, PIRLS, and IQB studies comparable with each other?

In principle, the achievement tests used in German large scale assessment studies (PISA, PIRLS, and IQB studies) correlate highly, but the underlying competence models differ. The IQB tests are based on the educational standards of the The Standing Conference of the Ministers of Education and Cultural Affairs of the Länder in the Federal Republic of Germany (Kultusministerkonferenz, KMK) and as a result more closely aligned with the German school curriculum than PISA tests.

The extent of comparability must be considered separately for reading and mathematical literacy and for secondary and primary education. Although it can be assumed that federal state differences can be well mapped using both measures, it is unfortunately not possible to analyse trends on a common metric.

The data sets of PISA 2012 and IQB National Assessment Study 2012 studies can be linked with each other using the ID variable [idstud_FDZ]. This allows you to compare correlations between the scaled test values of both studies.

Please note additionally:

  1. In contrast to the PISA surveys, reading and mathematical literacy are only tested together in the IQB study in primary school: Reading literacy was recorded in the IQB National Assessment Study 2009 (lower secondary level) and in the IQB National Assessment Study 2011 (primary school) as well as in the IQB Trends in Student Achievement 2015 (lower secondary level) and in the IQB Trends in Student Achievement 2016 (primary school). Mathematics competencies can be found in the IQB National Assessment Study 2012 (secondary level) and IQB National Assessment Study 2011 (primary level) as well as in the IQB Trends in Student Achievement 2016 (primary level) and the IQB Trends in Student Achievement 2018 (secondary level).
  2. If you wish to conduct analyses that include unpublished, novel comparisons between single federal states, our Rules of Procedure state that an extended application procedure with a review process applies.

Which PISA data can be linked to the IQB Studies?

The PISA 2012 datasets can be combined with the data from the National Assessment Study 2012. The student IDs in the data sets available at the FDZ at IQB have already been recoded in such a way that it is possible to link both data sources. Unfortunately, it is not possible to link the other PISA waves with the data of the IQB studies, as the ID variables cannot be recoded uniformly.

Is it possible to record the age of students (to the day) in the IQB National Assessment Studies/IQB Trends in Student Achievement Studies and in PISA?

Information on the year and age of birth of students is collected as standard in the IQB National Assessment Studies and PISA studies and is available for re- and secondary analyses of the data. For reasons of data protection, however, the exact date of birth was not recorded and is not available in the data sets. The exact test date is also not included in most data sets (in PISA 2009 this information is available). Frequently, however, the data sets contain an age variable that was formed using the year and month of birth in relation to the test date (e.g. in the IQB National Assessment Studies 2011, 2012 and in PISA 2012, 2009, 2006).

IQB National Assessment Study 2008/2009

Is there any information on the size of a town in the National Assessment Study 2008-9?

Unfortunately, the size of the location of the schools was not asked in the National Assessment Study 2008-9.

There are two separate weights in the French data set. Which is the correct one?

The National Assessment Study in French was carried out together with the standardization study in a joint survey in 2008 (see p. 8 in the Scale Manual for the study, which you can download PDF here (only German). The two samples for the standardization study and the National Assessment Study overlap by only about 1900 students. Therefore, separate weights are available for these samples. For a cross-country comparison sample, you should use the total student weight [wgtlvfr].

There is no kindergarten attendance variable in the data set for the National Assessment Study in French nor in the online scale book. Was this not collected?

Unfortunately, the attendance of kindergarten or pre-school was only recorded in the National Assessment Study in the German and English data sets.

Where can I find the student age in the data set for the National Assessment Study in French?

The month and year of birth of the students are included in the student dataset. The information is available both as teacher information [Tgebjah], [Tgebmo] and as student information [Sgebjah], [Sgebmo].

How do I carry out trend analyses with the National Assessment Study 2008/09 and the IQB Trends in Student Achievement 2015?

In the National Assessment Study 2008/09 achievement data in the subjects German, English and French were collected for the first time. The IQB Trends in Student Achievement 2015 is now the second survey on educational standards in these subjects at the end of 9th grade, making it possible to conduct trend analyses.

In order to depict trends, it is essential that the results from various surveys be presented on a uniform metric. For this purpose, the data from the National Assessment Study 2008/09 were transferred to the reporting metric of IQB Trends in Student Achievement 2015. The dataset available at the FDZ at the IQB for the National Assessment Study 2008/09 was updated accordingly. Furthermore, in the course of the evaluation and reporting on the IQB Trends in Student Achievement 2015, certain variables were newly created in order to be able to fall back on a uniform coding of the variables for both measurement dates (for example information on parents' occupations for the data from 2009 was also recoded according to ISCO-08 and the variables based on this were newly formed).

Please also note the notes on trend presentation in the Scale Manual for the PDF IQB Trends in Student Achievement Study 2015 (Ger/Eng) (in German).

A list of the new variables in the student dataset that are not included in the Scale Manual for the National Assessment Study 2008/9 can be found under Documentation.

IQB National Assessment Study 2011

What is the reason for missing values on the competence estimators (e.g. plausible values, competence levels, WLEs) in the National Assessment Study 2011?

Missing values on the competency estimators (e.g. plausible values, competency levels, WLEs) can be attributed to the fact that these children either did not participate in the competency test (no participation in the test booklet on test day 1 and/or test day 2, see variables [tr_t_th_tt1] and [tr_t_th_tt2] in the student dataset) or were excluded from the test for another reason (see variable ]tr_Ex" in the student dataset).

What are the WLE-reliabilities of the competency estimators and reference for scales of academic self-concept in the National Assessment Study 2011?

WLE Reliabilities:

  • - Reading = 0.669
  • - Listening = 0.605
  • - Orthography = 0.844
  • - Mathematics (global model) = 0.921

Reference for scales for academic self-concept:

  • Martin, M. O., & Preuschoff, C. (2008). Creating the TIMSS 2007 background indices. In J. F. Olson, M. O. Martin, & I. V. S. Mullis (Eds.), TIMSS 2007 Technical Report (pp. 281–338). Chestnut Hill, MA: TIMSS & PIRLS International Study Center, Boston College.

How do I link teachers’ and students’ data sets?

In the National Assessment Study 2011, one class per school was included in the sample (with a few exceptions in schools where students with special educational needs were not tested together with students without special educational needs).

The students data set and teachers data set can be combined by using the ID variable "idsch_FDZ". Before linking both data sets, double cases with identical "idsch_FDZ" (N=557) must be removed from the teachers’ data set. Duplicates result from the fact that the subjects German and Mathematics were taught by different teachers in some classes. In these classes, information is available from both teachers.

It is recommended to make the selection of data subject-specifically (variable names in the teacher data set: German: "LFa_a", mathematics: "LFa_f") should be selected.

How do I conduct trend analyses with the IQB Trends in Student Achievement 2015?

In order to carry out trend analyses between the National Assessment Study 2011 and the IQB Trends in Student Achievement 2016 and to replicate the Trends results, newly scaled plausible values are required for the competence area orthography.

  • New variables (N=15): SLvDOR01_Metrik2016 to SLvDOR15_Metrik2016

For the other competence areas (reading and listening) and mathematics, no adjustments to the plausible values are required to perform trend analyses. In order to replicate the results of the National Assessment Study 2011 for the competence area orthography, the original competence estimators should be used (variable names: SLvDOR01 to SLvDOR15).

IQB National Assessment Study 2012

The data sets of PISA 2012 and IQB National Assessment Study 2012 studies can be linked with each other using the ID variable [idstud_FDZ]. This allows you to compare correlations between the scaled test values of both studies.

What is the reliability of the scales BEFKI figural (wle.gff), C-test (wle.ctest) and highest ISEI of the family (HISEI)?

The WLE-reliabilities of the BEFKI (figural) and C-test scales are: BEFKI: 0.701; C-test: 0.884. Unfortunately, we cannot report a reliability coefficient for the highest ISEI in the family (HISEI), since this indicator was recorded via only one item.

How do I deal with missing values in multi-level models?

We recommend - as described in Chapter 10 of the report on the study LINK - the use of the Full Information Maximum Likelihood (FIML) approach to deal with missing values in multi-level models. IQB-analyses also used the PVs available in this dataset and only used FIML in Mplus afterwards.

Further methodological notes on dealing with missing values in multi-level models can be found in the following publications:

  • Grund, S., Lüdtke, O., & Robitzsch, A. (2018). Multiple imputation of missing data for multilevel models: Simulations and recommendations. Organizational Research Methods. doi: 10.1177/1094428117703686
  • Lüdtke, O., Robitzsch, A., & Grund, S. (2017). Multiple imputation of missing data in multilevel designs: A comparison of different strategies. Psychological Methods, 22, 141–165. doi: 10.1037/met0000096

Of the more than 44,000 students in the sample, about 40% have no performance data. What are the reasons for this?

The high percentage of missing values is due to the fact that not all students were presented with all competency tests, but a multiple matrix sampling was used. The missing values are therefore largely "Missing by Design". More information on test design can be found in the report on this study LINK (only german) in Chapter 4 and Chapter 13.

What are the new variables in the students` data set in version 4?

In the fourth version, new variables on learning time in biology, chemistry, physics and natural sciences have been added to the students` dataset. These are the number of learning hours per semester from the 5th to the 9th grade in each tested subject. In addition, the variables on the cumulative number of learning hours per week in the school years from grade 5 to grade 9 in each tested subject have been updated. Furthermore, the variables on the level of competence in mathematics, biology, chemistry and physics have been corrected.

Here you can find an overview of the new variables:

  • Learning time Biology: pstdbio051.r, pstdbio052.r, pstdbio061.r, pstdbio062.r, pstdbio071.r, pstdbio072.r, pstdbio081.r, pstdbio082.r, pstdbio091.r, pstdbio092.r
  • Learning time Chemistry: pstdche051.r, pstdche052.r, pstdche061.r, pstdche062.r, pstdche071.r, pstdche072.r, pstdche081.r, pstdche082.r, pstdche091.r, pstdche092.r
  • Learning time Physics: pstdphy051.r, pstdphy052.r, pstdphy061.r, pstdphy062.r, pstdphy071.r, pstdphy072.r, pstdphy081.r, pstdphy082.r, pstdphy091.r, pstdphy092.r
  • Learning time Natural Sciences: pstdnws051.r, pstdnws052.r, pstdnws061.r, pstdnws062.r, pstdnws071.r, pstdnws072.r, pstdnws081.r, pstdnws082.r, pstdnws091.r, pstdnws092.r

Here you can find an overview of the updated variables:

  • Learning time in biology, chemistry, and physics: lzbio, lzche und lzphy
  • Competency Levels: Mathematics (global scale): pv_1_GL_stufe - pv_15_GL_stufe
  • Competency Levels: Biology (content knowledge): pv_1_BF_stufe - pv_15_BF_stufe
  • Competency Levels: Biology (scientific inquiry): pv_1_BE_stufe - pv_15_BE_stufe
  • Competency Levels: Chemistry (content knowledge): pv_1_CF_stufe - pv_15_CF_stufe
  • Competency Levels: Chemistry (scientific inquiry): pv_1_CE_stufe - pv_15_CE_stufe
  • Competency Levels: Physics (content knowledge): pv_1_PF_stufe - pv_15_PF_stufe
  • Competency Levels: Physics (scientific inquiry): pv_1_PE_stufe - pv_15_PE_stufe

Is it possible to establish links between students and teachers via the link data set?

The linkability between student and teacher data sets is also limited when using the link data set (data set: IQB-LV-2012_Link_SFB_LFB_SLFB).

The information on the subjects taught in the link data set was collected by answering the following question in the teachers' questionnaire: "Do you teach the classes/courses in the subjects mathematics, biology, chemistry, physics or natural sciences?

The information in the link data set does not allow us to determine with certainty whether a teacher taught a specific student. Some students were taught by several of the teachers who were tested. It is also possible that a teacher taught a student in mathematics and also taught another tested class in physics. A clear assignment can be made using the variables for the course designation in the teacher data set (data set: IQB-LV-2012_LFB; variable names: luntflvteil01_1_FDZ to luntflvteil24_1_FDZ) and in the student data set (data set: IQB-LV-2012_SFB; variable names: tkursdiffdeu_FDZ, tkursdiffmat_FDZ, tkursdiffbio_FDZ, tkursdiffche_FDZ, tkursdiffphy_FDZ, tkursdiffnwi_FDZ).

Nevertheless, it is not possible for all students and teachers to clearly assign the variables. On this challenge, you can consult chapter 12 (especially subchapter 12.6) of the report on the study, which is available online. LINK (only german)

In chapter 12.6 (p. 381), it states:

"Für die Analysen zum Zusammenhang von Lehrbefähigung, Fortbildungsteilnahme und Schülerkompetenzen wurden zunächst die Angaben der Lehrkräfte mit den im IQB-Ländervergleich erreichten Kompetenzen der von ihnen unterrichteten Schülerinnen und Schüler verknüpft, wobei eine eindeutige Zuordnung für 41 Prozent der Schülerinnen und Schüler in Mathematik, für 39 Prozent in Biologie, für 35 Prozent in Chemie und für 47 Prozent in Physik vorgenommen werden konnte. Diese vergleichsweise geringe Zuordnungsquote von weniger als 50 Prozent lässt sich unter anderem darauf zurückführen, dass ein Teil der Lehrkräfte den Fragebogen nicht ausfüllte, bei Kursunterricht zum Teil Informationen über die Zuordnung zwischen Lehrkraft und Schülerinnen und Schülern fehlten sowie für einen Teil der Jugendlichen Angaben von zwei Lehrkräften desselben Faches innerhalb einer Klasse vorlagen. War letzteres der Fall, wurden die Daten der Schülerinnen und Schüler den Lehrkräften nicht zugeordnet, um Fehlzuordnungen zu vermeiden."

How are the data sets of the National Assessment Study 2012 linked to PISA 2012?

The data sets of the National Assessment Study 2012 can be linked with the data sets of the 9th graders in the PISA 2012 study, which is also available at the FDZ of the IQB. The classes of the PISA 2012 study (n = 9 998; two classes per school) participated in the competence testing of the IQB on the second day of testing. The linkage of the data sets is achieved via the variable idstud_FDZ.

Unfortunately, it is not possible to link the other PISA waves with the data of the IQB studies because the ID variables cannot be recoded uniformly.

IQB Trends in Student Achievement 2015

Does the IQB Trends in Student Achievement 2015 Study include variables on bullying?

No, not exactly. The students were asked about their well-being at school (scale name: Swas) (e.g. "I feel like an outsider in this school", "I feel uncomfortable and out of place in this school", "Other pupils seem to like me"), but not about bullying. Variables on this topic can be found in the PISA 2015 study.

Does a technical report on sampling and a scale manual on items exist for the IQB Trends in Student Achievement 2015 Study?

Information on sampling and test design can be found in the report volume (chapter 3 and chapter 11) and the Codebook IQB Trends in Student Achievement 2015 (Ger/Eng) (in German).

What scaling was used as a basis for the test score for BEFKI 8-10?

Weighted Likelihood Estimate Estimators (WLE, Warm, 1989) are individual estimators of abilities based on item-response models. WLEs can be interpreted in a similar way to z-values (average ability = 0, higher values represent higher abilities), but the standard deviation for WLEs is not 1. The WLEs were formed on the basis of the sample of the IQB Trends in Student Achievement, neither age nor age-group-related norm values were used.

More information on the interpretation of WLEs can be found in this publication:

How do I identify class groups using the IDTESTGROUP variable?

The variables [TR_Class] and [IDTESTGROUP_FDZ] have been emptied for data protection reasons. You can use the [idtestgroup] variable in JoSuA to identify clusters (this variable contains all information of the emptied variable [IDTESTGROUP_FDZ]). Additionally, you can use the variable [idteach_d_FDZ1] (German teacher): Each German teacher is usually assigned to a test group ([idtestgroup]).

How are the teachers’ and students’ data sets matched?

  • 1) data sets for the subjects German/English

For the matching of these data sets, the teacher-ID (German teachers „idteach_d_FDZ1“, English teachers „idteach_e_FDZ1“ or special school teachers „idteach_f_FDZ1“ ) can be used. Since the degree to which teachers are obliged to participate in the survey varies between the Länder, there may be systematically missing values amongst teachers’ data. At the same time, there are classes with two teachers for the same subject.

  • 2) data sets for the subject French

Matching can simply be done via the school ID. Exactly one class or course was tested within each school. For the school with the recoded ID 170 the special case arises that two teachers were teaching the same class and both were interviewed. Before matching this duplication must be handled, e.g. by excluding one case.

Why does an ID occur twice in the principal data set?

There is a duplicate ID in the principle data set (245). There is a duplicate of the principle questionnaire with IDSCH 538712, as the school was sent another questionnaire at the request of the principal. The data are mostly congruent; we would recommend data users to randomly choose one row for analyses.

What changes were made version 5?

  • Correction of the variable ‘school type’

In the fifth version of the IQB Trends in Student Achievement 2015, an incorrect coding of missing values on the variable school type takes effect. This variable now no longer has any missing values.

What changes were made version 4?

  • new data sets

The fourth version of the IQB Trends in Student Achievement 2015 contains three data sets (students’ data set, teachers’ data set, principals’ data set) with a sample of Länder where French is most commonly the first foreign language students learn. The results for this complementary sample are reported separately in the IQB Trends in Student Achievement 2015 report (https://www.iqb.hu-berlin.de/fdz/bt/BT2015).

  • new variables

Variables were recalculated and recoded in the German/English teachers’ data set in order to allow replications of the report on the IQB Trends in Student Achievement 2015 Study.

NEW: Lbfg-deu and Lbfg_eng RECODED: Lfremd_deu and Lfremd_eng

IQB Trends in Student Achievement 2016

How are teacher data set and student data sets linked?

In the IQB Trends in Student Achievement 2016 Study there is one ID (ZIDteach) in the student data set and there are several IDs in the teacher data set (see Scale Manual):

  • ZIDteachD = ID German teachers
  • ZIDteachM = ID Mathematics teachers
  • ZIDteachD2 = ID for German teachers who teach in several tested classes (only at special education schools)
  • ZIDteachM2 = ID for mathematics teachers who teach in several tested classes (only at special education schools)

For linkage the variable ZIDteach in the student data set would have to be renamed in order to match the teachers` data (it can be used as the variables ZIDteachD or ZIDteachM etc.). This is necessary in order to use only one teacher per student group in wide formatdata sets. The variable ZIDteach is a combination of school ID (IDSCH) and a string variable of the class taught at special education schools. This is necessary because at special education schools all students of the 4th grade and all German and mathematics teachers took part in the study.

This means that within a school there are several classes which cannot be identified in any other way. At general education schools, one class per school was tested. There is a maximum of two teachers per class (i.e. one German teacher and one maths teacher). Thus the variable ZIDteach at general education schools corresponds to the school ID (IDSCH).

What is the ‘linking error’ data set used for?

Every statistical estimate is connected with uncertainty, which can have various reasons. For large school performance studies, measurement error and sampling error play a particularly important role. When estimating how the skills achieved have changed between surveys, linking error must also be taken into account as a third possible source of uncertainty. When calculating trends, the respective linking error may need to be taken into account when determining the standard error.

A detailed description of this can be found in the technical chapter of the report on the IQB Trends in Student Achievement 2015 (Sachse et al., 2016).

Further methodological information is also contained in the R package eatRep.

IQB Trends in Student Achievement 2018

How are teacher data set and student data sets linked?

All data directly related to the students have already been merged into one data set. Thus, in addition to the information provided by the schools on the individual students, the results of the tests on competencies and basic cognitive skills and the survey of the students, this data set also contains the information from the parents' questionnaire. The only exception is the students' information on social networks, which is provided in a separate data set. The results of the teachers' survey and the survey of the school headmasters can also be found in separate data sets and can be merged with the students' data if necessary.

The data on social networks can be linked to the other data on the students via the student ID ("IDSTUD").

Information from the teacher survey is found in two separate data sets. On the one hand, there is a "general teacher data set", which contains information about the teachers themselves (e.g. their training). Each teacher who participated in the survey corresponds to one line in this data set. On the other hand, a "learning group-specific teacher data set" is provided, which contains information of the teachers on individual learning groups (e.g. on characteristics of their teaching in the respective learning group). In this data set, one row corresponds to one learning group, so that information in several rows can come from the same teacher. Both data sets can be linked with the help of the teacher ID (variable "IDTEACH"). The linking of teachers and students is done using the matching data set and the learning group specific data set based on the matching ID ("IDMATCHING"). This variable represents a unique identifier of a learning group at general schools and special schools.

The linking of the student data with the data of the principals can be done for both the general schools and the special schools via the school ID (variable "IDSCH").

What is the ‘linking error’ data set used for?

Every statistical estimate is connected with uncertainty, which can have various reasons. For large school performance studies, measurement error and sampling error play a particularly important role. When estimating how the skills achieved have changed between surveys, linking error must also be taken into account as a third possible source of uncertainty. When calculating trends, the respective linking error may need to be taken into account when determining the standard error.

A detailed description of this can be found in the technical chapter of the report on the IQB Trends in Student Achievement 2015 (Sachse et al., 2016).

Further methodological information is also contained in the R package eatRep.

IQB Trends in Student Achievement 2021

Are the competence estimators of the PISA, IGLU and IQB studies comparable with each other?

In principle, the tests from PISA and the IQB studies correlate highly, but the underlying competency models differ. The IQB tests are based on the educational standards of the The Standing Conference of the Ministers of Education and Cultural Affairs of the Länder in the Federal Republic of Germany and thus more closely aligned with the schools´ curriculum than the PISA tests. Comparability can be tested using IRT methods based on studies in which both PISA and IQB items were used. Some studies for comparison are for example

The extent of comparability must be considered separately for reading and mathematical literacy and for secondary and primary education. Although it can be assumed that federal state differences can be well mapped using both measures, it is unfortunately not possible to analyse absolute trends on a common metric.

The data sets of PISA 2012 and IQB National Assessment Study 2012 studies can be linked with each other using the ID variable [idstud_FDZ]. This allows you to compare correlations between the scaled test values of both studies.

Please note additionally:

  1. In contrast to the PISA surveys, reading and mathematical literacy are only tested together in the IQB study in primary school: Reading literacy was recorded in the IQB National Assessment Study 2009 (lower secondary level) and in the IQB National Assessment Study 2011 (primary school) as well as in the IQB Trends in Student Achievement 2015 (lower secondary level) and in the IQB Trends in Student Achievement 2016 (primary school). Mathematics competencies can be found in the IQB National Assessment Study 2012 (secondary level) and IQB National Assessment Study 2011 (primary level) as well as in the IQB Trends in Student Achievement 2016 (primary level) and the IQB Trends in Student Achievement 2018 (secondary level).
  2. If you wish to conduct analyses that include unpublished, novel comparisons between single federal states, our Rules of Procedure state that an extended application procedure with a review process applies.

Is it possible to record the age of students (to the day) in the IQB National Assessment Studies/IQB Trends in Student Achievement Studies and in PISA?

Information on the year and age of birth of students is collected as standard in the IQB National Assessment Studies and PISA studies and is available for re- and secondary analyses of the data. For reasons of data protection, however, the exact date of birth was not recorded and is not available in the data sets. The exact test date is also not included in most data sets (in PISA 2009 this information is available). Frequently, however, the data sets contain an age variable that was formed using the year and month of birth in relation to the test date (e.g. in the IQB National Assessment Studies 2011, 2012 and in PISA 2012, 2009, 2006).

Which PISA data can be linked to which IQB National Assessment Studies/IQB Trends in Student Achievement Studies?

The PISA 2012 data sets can be combined with the IQB National Assessment Study 2012. The students IDs have already been recoded in the data sets available at the FDZ at IQB in such a way that a linkage of both data sources is possible. Unfortunately, it is not possible to link the other PISA waves with the data from the IQB IQB National Assessment Studies /IQB Trends in Student Achievement Studies because the ID variables cannot be recoded uniformly.

At what levels was the Germany PISA data collected?

In the German PISA studies, information is only available at the federal state level. Please note that special conditions of use must be observed when analysing data from the federal states. You can read them here:

What is the number of classes drawn per school in PISA surveys?

Information on sampling in the studies can be found in the results reports or scale manuals. Here is a brief summary of sampling in PISA:

  • PISA 2000: random selection of 28 15-year-olds and 10 non-15-year-old ninth graders per school; thus, full classes were not drawn, analyses can only be done at the school level
  • PISA 2003: random selection of 15-year-olds per school; in addition, two complete 9th graders were drawn per school for the national expansion; in the PISA-E data, however, no class-based sampling was realised.
  • PISA 2006: school-based sampling, then random selection of 15-year-olds per school; at the schools in the international sample (PISA_I), students from two complete 9th grades were additionally tested
  • PISA 2009: school-based sampling, additionally students from two complete 9th grades per school were tested
  • PISA 2012: school-based sampling, additionally students from two complete 9th grades per school were tested
  • PISA 2015:school-based sampling, additionally random selection of 15 ninth graders per school
  • PISA 2018: school-based sampling, additionally random selection of 15 ninth graders per school

How many students in special and vocational schools are included in the PISA data?

Special needs and vocational students were covered separately in the above PISA surveys. The sample sizes for these subgroups are given below. They are based on the data in the German PISA Extended Samples (PISA-E) in the student and school management data sets. Where appropriate, there may be slight differences from the reported sample sizes in the results reports.

  • PISA 2000 E:
    • - 9th grade: n= 11 students in vocational schools, n= 22 students in special schools out of a total of n= 34,754 students
    • - 15-year-olds: n= 241 students at vocational schools, n= 799 students at special schools out of a total of n= 35,584 students
    • - School data set: n= 18 vocational schools, n= 4 special schools out of a total of n= 1,342 schools
  • PISA 2003 E (here no differentiation between data sets for 9th grade & 15-year-olds possible):
    • - 9th grade: n= 654 students at vocational schools, n= 1,712 students at special schools out of a total of n= 46,185 students
    • - School data set: n= 43 vocational schools, no special schools out of a total of n= 1,411 schools
  • PISA 2006 E:
    • - 9th grade: no students at vocational schools or special schools in the data set
    • - 15-year-olds: n= 625 students at vocational schools, n= 2,560 students at special schools out of a total of n= 39,573 students
    • - School data set: n= 42 vocational schools, no special schools out of a total of n= 1,496 schools
  • PISA 2009 E:
    • - 9th grade: no students at vocational schools or special schools in the data set out of a total of n= 9,461 students
    • - School data set: n= 9 vocational schools, n= 13 special schools out of a total of n= 226 schools
  • PISA 2012 E:
    • - 9th grade: no students at vocational schools, n= 153 at special schools out of a total of n= 9,998 students
    • - 15-year-olds: n= 99 students at vocational schools, n= 139 at special schools out of a total of n= 5,001 students
    • - School data set: n= 7 vocational schools, n= 12 special schools out of a total of n= 230 schools
  • PISA 2015 E:
    • - 9th grade:no students at vocational schools, n= 165 at special schools out of a total of n= 4,149 students
    • - 15-year-olds: n= 160 students at vocational schools, n= 134 at special schools out of a total of n= 6,504 students
    • - School data set: n= 8 vocational schools, n= 12special schools out of a total of n= 205 schools
  • PISA 2018 E:
    • - 9th grade: no students at vocational schools, n= 115 at special schools out of a total of n= 3,567 students
    • - 15-year-olds: n= 184 students at vocational schools, n= 98 at special schools out of a total of n= 5,451 students
    • - School data set: n= 10 vocational schools, n= 7 special schools out of a total of n= 191 schools

Can teacher and student data be linked in PISA?

Unfortunately, linking is only possible for the partial data sets of 9th graders (the data sets of 15-year-olds include cross-school samples). In most PISA waves, two 9th graders were drawn, but the partial data sets often lack a unique class ID.

Here is an overview in bullet points of the individual PISA waves:

  • - PISA 2000: no teacher questionnaire was used here.
  • - PISA 2003:
    • partial data set "PISA-I-9th grade": teacher questionnaires contain questions at school level, not at class level; a link via the variable [idclass_FDZ] is possible, but in the teacher data set there is a high proportion of missing values on this variable (presumably because many teachers were surveyed per school)
    • partial data set "PISA-E": no teacher questionnaires available
  • - PISA Plus 2003-2004: a linkage is possible in principle, but teacher data would have to be imported from PISA 2003 data and are only available at the first measurement point.
  • - PISA 2006
    • partial data set "PISA-E": no teacher data set for 9th grades available, linkage only possible at school level
    • partial data set "PISA-I": no clear linkage possible, as the teacher data set does not contain a class ID.
  • - PISA 2009: also no class ID in the teacher data set, but linking via idsch and variable [LF39M01] (German taught in PISA class: yes vs. no) partially possible; however, two 9th grades were drawn from each school.
  • - PISA 2012: Linking is possible in principle via class name variables (teacher data set: class_FDZ; student data set: ClassName_FDZ) but difficult to achieve in practice, as the metric of the school ID does not correspond between the two sub-data sets and there is a high proportion of missing values on class name variables (I interpret reports from PISA staff that linking is not successful in the majority of cases).
  • - PISA 2015: Linkage is not directly possible, as all teachers in the drawn schools were surveyed.
  • - PISA 2018: A link between teachers and students via the variable "TEACHCLASS_ID" is not possible until the end of 2022 due to a blocking notice. However, this variable also only contains the information whether the teacher has taught a ninth grade or not. This is because almost all teachers in the drawn school were surveyed. Alternatively, the variable "TEACHERID" can be used, but this variable also does not allow a clear assignment between students and the corresponding teacher.

For which PISA data is a repeated measures data set available?

A repeated measures data set is available for PISA-2003 (PISA-Plus 2003, 2004) and PISA-2012 (PISA-Plus 2012, 2013).

How were the science literacy tests developed in PISA?

In contrast to the IQB National Assessment Studies, the science tests in PISA are not curricularly anchored or subject-specifically designed. Therefore, there are no subtests for biology, physics and chemistry in PISA. Instead, PISA tests scientific literacy (see e.g. OECD, 2006). This involves skills that are significant in situations in which one is confronted with science and technology. These situations relate to physical systems, living systems, earth and space systems and technological systems. Specifically, the following competencies are tested:

(a) recognise scientific issues

b) describe, explain and predict scientific phenomena

c) use scientific evidence to make decisions.

More information on the concept and the test (including sample items) can be found here:

Whereas successful applicants for the data of PISA 2000, 2003 or 2006, respectively, will receive the German data from both the international study and the national supplementary study, this will not be the case with PISA 2009 and subsequent PISA waves. As of PISA 2009 the national supplementary study was replaced by the IQB Ländervergleich. The data from the IQB Ländervergleich will not automatically be provided together with the German data of the international PISA 2009 study, but is also available from the Research Data Centre (FDZ) at the Institute for Educational Quality Improvement (IQB) upon application. Further information on the IQB Ländervergleich in primary education can be found here, in secondary education here, and on our study pages.

PISA 2000

Please observe the following conditions for using the PISA 2000 data:

  1. Any publication using the German PISA 2000 data sets (PISA-I or PISA-E) must carry the following notice: "PISA 2000 was designed in Germany as a national research programme by the German PISA Consortium (Jürgen Baumert, Eckhard Klieme, Michael Neubrand, Manfred Prenzel, Ulrich Schiefele, Wolfgang Schneider, Klaus-Jürgen Tillmann, Manfred Weiß). It was lead-managed by Professor Dr. Jürgen Baumert, Max-Planck Institute for Human Development, Berlin. Primary research results have been published, e.g., in Baumert et al. (2001, 2002, 2003). Survey questionnaires have been documented in Kunter et al. (2002). We thank the German PISA Consortium and the Research Data Centre (Forschungsdatenzentrum, FDZ) in Berlin for granting permission to conduct this secondary analysis and for their support." The references cited in this footnote must be listed in your bibliography in the usual style.
  2. In line with parents' consent to having their children tested in the PISA 2000 Assessment, information on students' basic cognitive skills (KFT; kognitive Grundfähigkeiten) can only be used as covariates, not as target variables. The descriptive representation of distributions, especially the publication of group comparisons, as well as the use of the aggregate KFT test or of test parts as dependent variables is strictly prohibited. In the event of a violation of this agreement the publication concerned must be withdrawn, with its author acknowledging his or her infringement of the right of consent of the persons involved.
  3. As the contracting authorities and the Consortium have agreed to preserve the anonymity of the schools participating in the PISA Assessment, a ban has been placed on publishing any analyses in which individual schools can be identified or made identifiable.
  4. When the Federal Government and the Federal States awarded the contract for the PISA 2000 Assessment to the PISA 2000 Consortium, this was done with the proviso that no school-type related performance comparisons between Federal States shall be drawn. This agreement also applies to any and all secondary analyses.

There are different PV variables for mathematics in the 9th grade student data set. What is the difference between the PV variable groups?

  • - pvxnatm]: These PVs are the national mathematics tests and their results are reported on a national metric with a mean of 100 and a standard deviation of 30.
  • - [pvxnatmi]: These PVs are also based on the national test items, but are reported on the international metric with mean 500 and standard deviation 100 (therefore they correlate to one with the PVs [pvxnatm]).
  • - [pvxmg]: This is the combined performance score. All international and all national items are included in this score. According to the scale manual for the PISA 2000 study (Kunter et al., 2002; p. 77ff.; in the scale manual these items are labelled [NPV1MG1D]), it is recommended to use these PVs: "Results of various dimensional analyses show that it is appropriate to map the international and the national mathematics test on an overall dimension of basic mathematical literacy" (Kunter et al., 2002; p. 78). You can download the scale manual here ; if necessary, the supplements to the scale manual will also help you.

PISA 2003

Please observe the following conditions for using the PISA 2003 data:

  1. In accordance with the parents' consent to the PISA 2000 study, the information contained in the data sets on pupils' basic cognitive skills (KFT) may only be used as covariates, not as target variables. A descriptive presentation of distributions, in particular the publication of group comparisons and the use of the overall KFT test or parts of the test as dependent variables, is strictly prohibited. In the event of a breach of this agreement, the publication in question must be revoked with reference to the violation of the consent rights of the persons involved.
  2. Due to the commitment made by the funder and the consortium to maintain the anonymity of the participating schools, no analyses may be published in which individual schools are identifiable or can be made identifiable.

Were the teachers also interviewed again at the 2nd measurement point?

Yes, but unfortunately this information is not included in the data set. They are part of the COACTIV project, but unfortunately this data has not (yet) been made available for secondary analyses.

Are there variables that include the subjects taught?

The corresponding data for measurement time point (MZP) 1 can be taken from the teacher data set of the PISA 2003 study (data set name: PISA2003-I_Datensatz_Lehrkraft_9Kl) and transferred to the PISA I-Plus 2003-4 data set via the variable [idclass_FDZ]. However, the PISA 2003 teacher data set only contains information on whether any of the subjects taught are mathematics ([fama_all]), German ([fadeu_al]), physics ([faphy_al]), biology ([fabio_al]) or chemistry ([fache_al) (1 = yes).

Does the variable [fama_all] include the indication whether it is a mathematics teacher? If yes, which specification (0/1)? Is it possible to find out the associated subject-specific final grade (state examination/main examination)?


The variable [fama_all] indicates whether the teacher teaches mathematics (1 = yes). Unfortunately, it is not possible to find out which final grade corresponds to the subject mathematics on the basis of the data set available at the FDZ at IQB.

Is there a variable for the parents' occupation?

The following grouped information on the parents' occupational status is available in the teacher dataset of the PISA 2003 study:

  • bmut11_1: Mother's occupational status: civil servant (1 = yes)
  • bmut12_1: Mother's occupational status: employee (1 = yes)
  • bmut13_1: Mother's occupational status: self-employed (1 = yes)
  • bvat11_1: Father's occupational status: civil servant (1 = yes)
  • bvat12_1: Father's occupational status: employee (1 = yes)
  • bvat13_1: Father's occupational status: self-employed (1 = yes)

The ISCO codes (grouping of occupations) of the parents are also included in the dataset:

  • iscolfm1: ISCO code for mother's occupation
  • iscolff1: ISCO code for father's occupation

PISA-I-Plus 2003-4

No separate scaling manual exists for PISA-I-Plus 2003-4. The student questionnaire used the same scales as in the PISA 2003 study, so the corresponding documentation can be used. Further information can be found on our page about the PISA 2003 study and on the website of the IPN-Leibniz Institute for Science and Mathematics Education.

On the PISA 2003 study website, you can view excerpts from the book mentioned above, courtesy of Waxmann-Verlag:

  • Prenzel, M. und Deutsches PISA-Konsortium (2006). PISA 2003: Untersuchungen zur Kompetenzentwicklung im Verlauf eines Schuljahres. Münster u. a.: Waxmann.

PISA 2009

Why are there no sub-data sets for 15-year-olds and 9th graders in the PISA 2009 study as in the other PISA studies?

Unfortunately, we have only received the PISA 2009 data for ninth graders from the data-providing consortium. The PISA 2009 data for 15-year-olds are freely available on the OECD website.

When was the PISA 2009 test conducted in Germany?

In Germany, the PISA 2009 survey was conducted in April and May 2009; further information on the implementation can be found in the PDFResults Report on PISA 2009 in Germany (p. 16ff.).

PISA 2012

Is there a possibility to get the federal state identifier in PISA 2012?

The federal state identifier is included in the data sets. We will gladly make this data set version available to you via the controlled remote computer access JoSuA. However, for this we would have to conclude an additional agreement with you on the use of this data version. Please contact us for this purpose. The data sets of the PISA 2015 study also contain the federal state identifier.

PISA Plus 2012-2013

In PISA Plus 2012-13, is it possible to link not only students but also principals and teachers longitudinally?

Unfortunately, the teacher IDs cannot be recoded uniformly between the two measurement points, so that it cannot be ensured whether the same teacher completed the questionnaire at both measurement points. This also applies to the principals, but at least the school IDs can be linked between the two measurement points. Thus, information about the participating schools can be linked longitudinally. However, the number of items used in both surveys is relatively small.

There is no separate scaling manual for PISA Plus 2012-2013. The same scales were used in the survey as in the PISA 2012 study, so the corresponding documentation can be used. Further information can be found on our page about the PISA 2012 study and on the website of the Center for International Educational Comparative Studies (ZIB).

PISA 2015

What do the endings "TA" and "NA" in the variable names of the student questionnaire in PISA 2015 stand for?

The endings in the student questionnaire indicate the following:

  • "TA": This is a trend analysis item that has already run in this form at least once in a previous PISA wave.
  • "NA": This suffix indicates that this item has never been used in a previous PISA cycle.

More information on item naming can be found PDF here (e.g. footnote 4) and in the PISA 2015 Technical Report.

Can teacher and student data sets be linked in PISA 2015?

Unfortunately, it is not possible to link exactly which teacher teaches which student. However, a link at school level is possible.

PISA 2022

Are the competence estimators of the PISA, PIRLS and IQB studies comparable with each other?

In principle, the achievement tests used in German large scale assessment studies (PISA, PIRLS and IQB studies) correlate highly, but the underlying competence models differ. The IQB tests are based on the educational standards of the The Standing Conference of the Ministers of Education and Cultural Affairs of the Länder in the Federal Republic of Germany (Kultusministerkonferenz, KMK) and as a result more closely aligned with the German school curriculum than PISA tests.

The extent of comparability must be considered separately for reading and mathematical literacy and for secondary and primary education. Although it can be assumed that federal state differences can be well mapped using both measures, it is unfortunately not possible to analyse trends on a common metric.

What else needs to be considered when analysing data?

More detailed methodological information on using data from large-scale evaluation surveys can be found in the eatRep tutorial (English).

PIRLS 2001

What is the difference between the two kinds of competence estimators in PIRLS 2001?

The PV sets [linkmatX] and [linknwX] are transformed competence scores, which are intended to link students` performance to the 1995 TIMSS study. The linking is based on items used in TIMSS 1995 as well as in IGLU 2001. Further information can be found in the results reports of IGLU 2001 (Bos et al., 2003, 2004, 2005).

There is no information on grades in the areas of German, grammar, reading, spelling, mathematics and general knowledge (zendtsch, grammar, reading, rechtsch, zenmathe, zensk) in the scale manual. How was this information collected?

The grades were recorded via the student participation list. Grades with decimal places may result from the fact that in some federal states or schools, partial grades (e.g. geometry, arithmetic) are assigned which are then averaged to one grade. In addition, in some federal states, partial grades are given in German for spelling, grammar and reading, which are also available in the data set (since these separate grades for sub-areas in German are not awarded in every state or school, the missing parts on these variables are comparatively high). In the results report on IGLU 2001 (Bos et al., 2004), Chapter IX (p. 191; "School career recommendations by teachers for children at the end of the fourth grade") briefly discusses the sub-grades.

Where can I find information on the test scores at federal state level?

Information on the achieved test scores at federal state level can be found in the following report:

  • Bos, W., Lankes, E.-M., Prenzel, P., Schwippert, K., Valtin, R. & Walther, G. (Hrsg.) (2004). IGLU : einige Länder der Bundesrepublik Deutschland im nationalen und internationalen Vergleich. Münster u.a.: Waxmann.

Why is the sample size larger than in later PIRLS surveys?

In 2001, the states of Baden-Wuerttemberg, Bavaria, Brandenburg, Bremen, Hesse, North Rhine-Westphalia, and Thuringia decided to expand the sample size in order to obtain representative results (cf. Bos et al. 2004a). However, since the sample in Thuringia did not meet the requirements of a random sample, only the other six states are compared in the comparative analyses by Bos and colleagues. Since 2011, the representative state comparison of the reading skills of primary school children has instead been carried out by the IQB.

PIRLS 2006

Why is the sample size larger than in later PIRLS surveys?

In 2006, all federal states decided to expand the sample size in order to obtain representative results for the individual states. Since 2011, the representative state comparison of primary school children's reading skills has instead been carried out by the IQB.

PIRLS 2011

Is a federal state identifier available in the data sets?

This information was not passed on to the data consortium by the collecting institute. For this reason, it is not possible to retrospectively import federal state identifiers to the data sets available at the FDZ at IQB.

How many special education schools are part of the sample and where can I find the school variable?

There is a variable in the school management data set that indicates the type of school: [ACNG32]. In this variable the category "3 = Other" includes special education schools. In addition, the missing category "9 - omitted or invalid" also includes two special schools. If you do not want to include special schools in your analyses, you can use the variable [ACNG32] and exclude categories 3 and 9. In total there are n=5 special education schools in the school management data set.

Why is the proportion of system-defined missing values for variables from the parents' questionnaire so high?

The proportion of system-defined missing values refers to the non-participation of a part of the students´ parents. A total of n=786 parents (19.7%) did not complete the questionnaire. In the results report (Bos et al., 2012, p. 55), this is pointed out as follows: "The response rate of the parents' questionnaires is 80 percent". Similar response rates can be found in other school achievement studies, especially if participation in the parent questionnaires was voluntary.

PIRLS 2016

How can I link the partial data sets?

A linked teacher-student data set already exists. There is also a data set containing only the teacher data. However, the teacher dataset does not allow for statements that are representative on a national level. There are 79 duplicate cases in the teacher-student data set. These are automatically taken into account when weighted analyses are performed. Linking the teacher-student data set to the tracking data set by student ID is only possible after the duplicate cases have been treated (e.g. exclusion, random selection or combination).

Why do the numbers of students differ between the student-parent data set and the tracking data set?

The tracking data set contains all students who were included in the sample. In the student-parent data set, on the other hand, only those who actually took part in the test are included. Non-participation can be explained by various reasons, which are summarised in the tracking data set in the variable "TR_EXCLUSION_FDZ". The sample size of n = 3959 in the student-parent data set is also used in the results report, so it is recommended to work with this sample size.

How was the age variable calculated?

The reference time for the age calculation (variable "ASDAGE") is the test date, which has been emptied due to data protection. If you have questions regarding the age at a different point in time, please contact us. We will be happy to advise you on how to use our data in accordance with data protection regulations.

Scales on classroom management: What is provided in the data and reports?

The student-parent data (SEFB) and the teacher-student data (LSFB) each contain four assessment scales on classroom management in the broader sense (classroom management/discipline (CM), cognitive activation (CA), social climate/support from the teacher (SC) and structuring (STR)). The constituent items were asked of the students and teachers, respectively, and are documented in the respective sections of the scale manual.

A sum score of the recoded items was formed for each scale. The coding instructions are described in chapter 9 of the results report (Stahns et al., 2017, pp. 264). The (four-level) items were first dichotomised by assigning 1 as the new value to the responses 3 and 4 in most cases. The other two responses were assigned the value 0. If the response to an item was missing because it was either omitted (missing by omission) or answered invalidly (missing invalid response), 0 was also assigned as the new value. A sum score was then calculated for the items of the scale.

Only people who had not been presented with the questions on classroom management were given empty/missing values on the individual items and sum scores. This means there is a valid sum score for every person who had the opportunity to answer the questions. However, this also means that people without valid answers to the individual items still have a valid sum score of 0. This applies to between 80 and 200 cases per scale.

What does this mean for data users?

The sum scores calculated in this way are included in the data and were the basis for the analyses in the IGLU 2016 report. To reproduce these results, the available sum scores have to be used. Data users who wish to conduct secondary analyses and use the constructs for classroom management in a different coding can use the individual unchanged items also included in the data to develop their own coding rules. To facilitate this, we provide an overview of the items, their assignment to the scales and the coding rules of the report chapter: Table (in German)

  • Stahns, R., Rieser, S. & Lankes, E.-M. (2017). Unterrichtsführung, Sozialklima und kognitive Aktivierung im Deutschunterricht in vierten Klassen. In A. Hußmann, H. Wendt, W. Bos, A. Bremerich-Vos, D. Kasper, E.-M. Lankes et al. (Hrsg.), IGLU 2016. Lesekompetenzen von Grundschulkindern in Deutschland im internationalen Vergleich (S. 251–277). Münster: Waxmann. verfügbar unter https://www.waxmann.com/index.php?eID=download&buchnr=3700